the best of a hundred nulls looks like this

A gate that asks whether a result is bigger than nothing is asking the wrong question when the result was chosen as the largest of many. The maximum of a set of random draws is not zero and is not small: it sits predictably some distance above the middle, and the distance grows with the number of draws. So the correct comparison is against the expected maximum of the set the winner was drawn from, and the size of that set is part of the result.

Close to a hundred configurations were swept, and exactly one survived every check it was given. It passed the engine check, the re-measurement against realized outcomes, the mirror comparison and every fold, which is precisely the situation in which a programme is most likely to announce something.

The rejection came from the arithmetic above. The survivor's statistic equalled the expected maximum of the nulls at the precision the comparison was made at — not near it, equal to it. That is what the largest of a hundred worthless configurations looks like, and nothing more. Its statistic also sat below the bound implied by the number of attempts. It went to review on those grounds and the review closed it.

The cost is visible in the shape of the record. This was the last configuration standing in a wave that had already closed four of its five families, and closing it emptied the wave.

What it proves is that the number of attempts is treated as part of the evidence.